154
Omar Boucelma, Mehdi Essid, and Yassine Lassoued
is a global key. Element Road has a mapping in R 3 , but its quality is not that good
because the spatial covers do not overlap. Consequently, we can reject this mapping.
Since it is not useful anymore to use this rule for the rest of the elementary queries,
we can prune it.
In the same way, we proceed with rest of the elementary queries without taking
into account the pruned branches of the mapping rules. Figure 7.14 depicts, respectively in white and gray, the branches that are kept and those that are pruned: black
cells show the quality values that are violated.
Elementary Query Reformulation
Once the ESQs evaluations are performed, it is possible to reformulate them in terms
of the source schemas. This is done in replacing subquery elements by their correspondents while adding the restrictions resulting from the mappings. For each subquery evaluation, there corresponds a reformulated query having the same quality
assertion as the evaluation.
Sectoral Execution Plans
Each ESQ of a query Q may involve a set of different data sources that play a certain
role in different regions of the space, see, for instance, example of rectangle B
of
Fig. 7.12. This leads to the creation of different sectors, each of them generating a
different way for evaluating subqueries. Sector partitioning is motivated by two main
reasons: (i) it avoids useless computations (while trying to join information of an
area to those of a disjoined area) and (ii) it allows one to estimate the quality of the
final execution plan. For each region, we compute a SEP, together with its quality
assertion.
Final Execution Plan
The FEP of a query Q corresponds to the selection of an nSEP for each sector. The
goal of the query rewriting process is to obtain a FEP with a good quality, that is each
of its SEPs has a good quality. Our goal is to choose, among the reformulated queries,
those that will participate in the FEP in a way that guarantees a satisfying quality.
Queries are ranked according to their qualities, in order to improve the quality of the
result, as well as the process speed. Making such a choice is difficult because quality
parameters are often contradictory. For example, improving the deficit rate means
degrading the surplus rate and the logical consistency. Thus, the problem is to obtain
a compromise leading to a rather complete result set without affecting its surplus rate
or its logical consistency.
7.7.3 Query Execution
Once the FEP is generated, it is ready to be executed. Each SEP is considered separately and its reformulated queries are extended with a spatial restriction over the
Omar Boucelma, Mehdi Essid, and Yassine Lassoued
is a global key. Element Road has a mapping in R 3 , but its quality is not that good
because the spatial covers do not overlap. Consequently, we can reject this mapping.
Since it is not useful anymore to use this rule for the rest of the elementary queries,
we can prune it.
In the same way, we proceed with rest of the elementary queries without taking
into account the pruned branches of the mapping rules. Figure 7.14 depicts, respectively in white and gray, the branches that are kept and those that are pruned: black
cells show the quality values that are violated.
Elementary Query Reformulation
Once the ESQs evaluations are performed, it is possible to reformulate them in terms
of the source schemas. This is done in replacing subquery elements by their correspondents while adding the restrictions resulting from the mappings. For each subquery evaluation, there corresponds a reformulated query having the same quality
assertion as the evaluation.
Sectoral Execution Plans
Each ESQ of a query Q may involve a set of different data sources that play a certain
role in different regions of the space, see, for instance, example of rectangle B
of
Fig. 7.12. This leads to the creation of different sectors, each of them generating a
different way for evaluating subqueries. Sector partitioning is motivated by two main
reasons: (i) it avoids useless computations (while trying to join information of an
area to those of a disjoined area) and (ii) it allows one to estimate the quality of the
final execution plan. For each region, we compute a SEP, together with its quality
assertion.
Final Execution Plan
The FEP of a query Q corresponds to the selection of an nSEP for each sector. The
goal of the query rewriting process is to obtain a FEP with a good quality, that is each
of its SEPs has a good quality. Our goal is to choose, among the reformulated queries,
those that will participate in the FEP in a way that guarantees a satisfying quality.
Queries are ranked according to their qualities, in order to improve the quality of the
result, as well as the process speed. Making such a choice is difficult because quality
parameters are often contradictory. For example, improving the deficit rate means
degrading the surplus rate and the logical consistency. Thus, the problem is to obtain
a compromise leading to a rather complete result set without affecting its surplus rate
or its logical consistency.
7.7.3 Query Execution
Once the FEP is generated, it is ready to be executed. Each SEP is considered separately and its reformulated queries are extended with a spatial restriction over the